Kun Zhao^1,2,3, Yanhui Ding³, Ying Han^4,17,18,19, Yong Fan⁵, Aaron F. Alexander-Bloch⁶, Tong Han⁷, Dan Jin^1,8, Bing Liu^1,8,9, Jie Lu¹⁰, Chengyuan Song¹¹, Pan Wang^12,13, Dawei Wang¹⁴, Qing Wang¹⁴, Kaibin Xu¹, Hongwei Yang¹⁰, Hongxiang Yao¹⁵, Yuanjie Zheng³, Chunshui Yu¹⁶, Bo Zhou¹³, Xinqing Zhang⁴, Yuying Zhou¹², Tianzi Jiang^1,8,9, Xi Zhang¹³*, Yong Liu^1,8,9*

¹Brainnetome Center & National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing, China;

²School of Biological Science and Medical Engineering, Beihang University, Beijing, China;

³School of Information Science and Engineering, Shandong Normal University, Ji’nan, China;

⁴Department of Neurology, Xuanwu Hospital of Capital Medical University, Beijing, China;

⁵Department of Radiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, USA;

⁶Department of Psychiatry, Yale University School of Medicine, New Haven, CT, USA;

⁷Department of Radiology, Tianjin Huanhu Hospital, Tianjin, China;

⁸School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China;

⁹Center for Excellence in Brain Science and Intelligence Technology, Institute of Automation, Chinese Academy of Sciences, Beijing, China;

¹⁰Department of Radiology, Xuanwu Hospital of Capital Medical University, Beijing, China;

¹¹Department of Neurology, Qilu Hospital of Shandong University, Ji’nan, China;

¹²Department of Neurology, Tianjin Huanhu Hospital, Tianjin, China;

¹³Department of Neurology, The Secondary Medical Center, National Clinical Research Center for Geriatric Disease, Chinese PLA General Hospital, Beijing, China;

¹⁴Department of Radiology, Qilu Hospital of Shandong University, Ji’nan, China;

¹⁵Department of Radiology, The Secondary Medical Center, National Clinical Research Center for Geriatric Disease, Chinese PLA General Hospital, Beijing, China;
¹⁶Department of Radiology, Tianjin Medical University General Hospital, Tianjin, China;

¹⁷Center of Alzheimer’s Disease, Beijing Institute for Brain Disorders, Beijing, China;

¹⁸Beijing Institute of Geriatrics, Beijing, China;19National Clinical Research Center for Geriatric Disorders, Beijing, China;

Abstract
Hippocampal morphological change is one of the main hallmarks of Alzheimer's disease (AD). However, whether hippocampal radiomic features are robust as predictors of progression from mild cognitive impairment (MCI) to AD dementia and whether these features provide any neurobiological foundation remains unclear. The primary aim of this study was to verify whether hippocampal radiomic features can serve as robust MRI markers for AD. Multivariate classifier-based SVM analysis provided individual-level predictions for distinguishing AD patients (n=261) from normal controls (NCs; n=231) with an accuracy of 88.21% and intersite cross-validation. Further analyses of a large, independent ADNI dataset (n=1228) reinforced these findings. In MCI groups, a systemic analysis demonstrated that the identified features were significantly associated with clinical features (e.g., apolipoprotein E (APOE) genotype, polygenic risk scores, cerebrospinal fluid (CSF) Aβ, CSF Tau), and longitudinal changes in cognition ability; more importantly, the radiomic features had a consistently altered pattern with changes in the MMSE scores over 5 years of follow-up. These comprehensive results suggest that hippocampal radiomic features can serve as robust biomarkers for clinical application in AD/MCI, and further provide evidence for predicting whether an MCI subject would convert to AD based on the radiomics of the hippocampus. The results of this study are expected to have a substantial impact on the early diagnosis of AD/MCI.

Key words: Hippocampal radiomic features; Multisite Alzheimer’s disease MRI; Independent cross-validation; Brain biomarker; Biological basis

Figure 1. Schematic of the data analysis pipeline. (a) Automatic segmentation of the hippocampus in the individual space with the LLL method. (b) Computation of radiomic features. (c) T-test between NC and AD and mega-analysis to identify differences at the multicenter level. (d) Classification analysis between AD and NC using intersite and intrasite crossvalidation to test the generalizability of the results. (e) Correlation between MMSE and radiomic features (and the decision value of the subject to the classification plane). (f) Validation of the ADNI dataset for using ADNI as training data with in-house data as testing data as well as using in-house as training data with ADNI data as testing data to test the generalizability of the results. (g) Relationship between radiomic features and clinical information (e.g., cognitive ability, APOE genetics, Aβ, Tau, and longitudinal changes in the cognitive abilities in MCI).

Article Link: https://www.sciencedirect.com/science/article/pii/S2095927320302140